Feast and Famine in Plant Genomes

Feast and famine in plant genomes

Jonathan F. Wendel1, Richard C. Cronn1, J. Spencer Johnston2 & H. James Price3 1Department of Botany, Iowa State University, Ames, IA, 50011, USA (Phone: +1-515-294-7172; Fax: +515 294-1337; E-mail: [email protected]); 2Department of Entomology, 3Department of Soil and Crop Sciences, Texas A & M University, College Station, TX 77843, USA

Key words: cotton, DNA content, genome evolution, genome size, Gossypium, Malvaceae, molecular evolution, 2C values

Abstract Plant genomes vary over several orders of magnitude in size, even among closely related species, yet the origin, genesis and significance of this variation are not clear. Because DNA content varies over a sevenfold range among diploid species in the cotton genus (Gossypium) and its allies, this group offers opportunities for exploring patterns and mechanisms of genome size evolution. For example, the question has been raised whether plant genomes have a ‘one-way ticket to genomic obesity’, as a consequence of retroelement accumulation. Few empirical studies directly address this possibility, although it is consistent with recent insights gleaned from evolutionary genomic investigations. We used a phylogenetic approach to evaluate the directionality of genome size evolution among Gos sypium species and their relatives in the cotton tribe (Gossypieae, Malvaceae). Our results suggest that both DNA content increase and decrease have occurred repeatedly during evolution. In contrast to a model of unidirectional genome size change, the frequency of inferred genome size contraction exceeded that of expansion. In conjunction with other evidence, this finding highlights the dynamic nature of plant genome size evolution, and suggests that poorly understood genomic contraction mechanisms operate on a more extensive scale that previously recognized. Moreover, the research sets the stage for fine-scale analysis of the evolutionary dynamics and directionality of change for the full spectrum of genomic constituents.

Introduction massive accumulation of retroelements, as elegantly shown for some species in the grass family (Bennet Plant genomes exhibit extraordinary variation in size, zen, 1996; SanMiguel et al., 1996; Bennetzen, 2000). ranging from approximately 125 Mbp in Arabidopsis The latter observations led naturally to a proposal that thaliana, known with exquisite precision now that genome size evolution in plants may largely be uni the genome has been nearly completely sequenced directional, from small to large (Bennetzen & Kellogg, (The Arabidopsis Genome Initiative, 2000), to over 1997a), with an overall pattern of ratcheting upward 120,000 Mbp in some relatives of lilies (Bennett & due to the combined effects of polyploidization and Leitch, 1995; Bennett et al., 1997; Leitch et al., 1998). retroelement accumulation. Because this remarkable range in DNA content is not The suggestion that plants may have a ‘one-way associated with variation in the basic complement of ticket to genomic obesity’ (Bennetzen & Kellogg, genes required for growth and development, it has 1997a) provided an appealing scenario that appeared become renowned as the ‘C-value paradox’ (Thomas, to account, at least in part, for the C-value paradox. 1971). Mechanistic explanations for the observed size Plants with small genomes are suggested to have been variation include repeated cycles of polyploidy over relatively effective in suppressing retrotransposon evolutionary time (Leitch & Bennett, 1997; Soltis & activity; whereas those with large genomes achieved Soltis, 1999; Otto & Whitton, 2000; Wendel, 2000), this condition by release from suppression (or new leading to progressively larger genomes, as well as invasion) and massive retroelement amplification. Be 38 cause the generality of this proposal is unknown the tribe using exemplar sampling, filled in gaps in and because alternative explanations exist, the sub our knowledge of chromosome numbers to ensure that ject has attracted considerable attention (Bennetzen & only diploids were included in the study, and em Kellogg, 1997b; Petrov, 1997; Leitch et al., 1998; ployed analytical methods for inferring ancestral con Vicient et al., 1999; Federoff, 2000; Rabinowicz, ditions for quantitative varying characters (Maddison, 2000; Shirasu et al., 2000; Petrov, 2001). The hy 1991; Martins & Hansen, 1997; Cunningham et al., pothesis of unidirectional genome size change has 1998; Martins, 1999; Oakley & Cunningham, 2000). been rigorously tested in only a few cases (Bennetzen & Kellogg, 1997a; Cox et al., 1998; Leitch et al., Materials and methods 1998). In addition, accumulating evidence suggests that genome size contraction may also be a com Organismal context and taxon sampling mon evolutionary occurrence. Supporting this sugges tion is recent evidence that underscores the diversity Included in the eight genera of the Gossypieae are four and scope of genomic deletional mechanisms (Petrov small genera with restricted geographic distributions et al., 1996; Petrov, 1997; Vicient et al., 1999; Kirik (Fryxell, 1979). Lebronnecia Fosberg, from the Mar et al., 2000; Petrov et al., 2000; Shirasu et al., 2000; quesas Islands, is a monotypic genus, as is Cephalo Petrov, 2001), the phylogenetically widespread dis hibiscus Ulbrich, from New Guinea and the Solomon tribution of plants with small genomes (Bennett & Islands. Gossypioides Skovsted ex J.B. Hutchinson Leitch, 1995; Bennett et al., 1997; Leitch et al., 1998), contains two species from East Africa and Madagas and the prevalence of high levels of infraspecific DNA car. Kokia Lewton, endemic to the Hawaiian Islands, content variation (Price, 1988; Kalendar et al., 2000) includes three extant and one extinct species. In ad that are not easily reconciled with an assumption of dition to these four small genera, the tribe includes unidirectional evolutionary change. four moderately sized genera with broader geographic A prerequisite for any discussion of the direction ranges: Hampea Schlechtendal comprises 21 neo ality of genome size change is that the phylogenetic tropical species; Cienfuegosia Cavanilles includes 25 relationships of the organisms under study are un species with an aggregate range that includes the neo derstood. Phylogenetic analysis provides an essential tropics and parts of Africa; 17 species are recognized framework for inferences of ancestral genome sizes in the pantropical genus Thespesia Solander; and Gos (Bennetzen & Kellogg, 1997a) as well as for polar sypium, as noted above, is the largest genus in the tribe izing genomic changes for specific homologous gen with about 50 species. omic regions. Hence, robust phylogenies are essential An initial phylogenetic scaffolding for the tribe for both global (genome size) and local (fine-scale) was erected several years ago based on sequences inferences of past genomic evolutionary events. A from the chloroplast gene ndhF and from nuclear particularly useful model in this respect is the group ribosomal ITS sequences (Seelanan et al., 1997). of plants comprising the cotton genus (Gossypium These analyses provided unequivocal support for L.) and its relatives in the small Malvaceous tribe several relationships among genera but left other as Gossypieae Alefeld. This monophyletic (LaDuke & pects of their evolutionary history unclear. Partic Doebley, 1995) tribe includes eight genera (Fryxell, ularly strongly supported was the close relationship 1979), the largest being Gossypium with approxim between Gossypioides and Kokia and their collec ately 50 species (Fryxell, 1992). Remarkable genome tive sister-taxon relationship to Gossypium. Thespesia size variation exists among the 45 diploid species of was implicated to be at least biphyletic and perhaps Gossypium, with 2C DNA contents ranging from ap polyphyletic, with a portion being sister to Lebron proximately 2.0 to 7.0 pg (Wendel et al., 1999). This necia and perhaps Hampea and the remainder be information motivated an exploration of genome sizes ing cladistically unresolved along with Cienfuegosia among its relatives, as well as phylogenetic relation and the (Hampea + Lebronnecia + Thespesia- in part) ships among the genera in the tribe. Our intention clade. Cephalohibiscus was not included in these was to generate the information required to understand analyses as it was not available. This earlier work the genome size variation that exists among diploid provided some insights into the evolutionary history species within this single tribe, and thereby facilitate of the group, but left unanswered several key questions inferences of the directionality of genome size change. of branching order, particularly as regards the earliest To accomplish this we estimated the phylogeny of divergences within the tribe. 39

Table 1. Species in the Gossypieae used for phylogenetic analysis, chromosome counts, and genome size measurementsa

Taxon Chromosome Genome no. (2n) size(2C)

C. hitchcockii (Ulbrich ex Kearney) Blanchard 20 2.3 C. tripartita H.B.K. Gürke 20 1.9 C. yucatanensis Millspaugh 22 2.0 G. kirkii (Mast.) J.B. Hutchinson 24 1.2 G. herbaceum L. 26 3.7 G. raimondii Ulbrich 26 2.0 H. appendiculata (J. Donnell-Smith) Standley 26 5.9 K. drynarioides (Seemann) Lewton 24 1.2 L. kokioides Fosberg 26 3.6 T. lampas (Cavanilles) Dalzell ex Dalzell and Gibson 26 3.2 T. populnea (L.) Solander ex Correa 26 8.2 T. thespesioides (R. Brown ex Bentham) Fryxell 26 3.2 M. sylvestris L. 42 3.0

aAll species were included in the molecular phylogenetic analysis with the exception of C. hitchcockii and C. yucatanensis. Chromosome counts determined in this study are in italics; others are from the literature. Genome sizes were estimated in the present study, except for the outgroup species M. sylvestris, which is from Bennett et al. (2000).

Chromosome numbers have been reported for six work as a model for studies of genome size evolution, of the eight genera (all except Cephalohibiscus and we conducted molecular phylogenetic analysis, made Lebronnecia). As summarized in Fryxell (1979), most new chromosome counts to confirm diploidy and fill species are diploid, with somatic numbers ranging in gaps in the data, and measured genome sizes out from 2n = 20 or 22 (Cienfuegosia)to 2n = 24 or 26 side of the genus Gossypium. The species evaluated (other genera). Polyploidy is unknown in the tribe (Table 1) in the molecular phylogenetic analysis in outside of one species of Thespesia (T. populneoides) cluded one representative of each of the available (Miege & Josserand, 1972) and five species of Gos genera in the tribe, except that two species of Gos sypium; the latter, in fact, provide one of the clas sypium were sampled, which vary twofold in DNA sic models of allopolyploid evolution (Wendel et al., content, and three species of Thespesia were sampled 1999; Wendel, 2000). DNA content has been determ because of their variable 2C DNA content and pre ined only for members of Gossypium, where a rich his vious suggestions of generic paraphyly or polyphyly. tory of genetic and cytogenetic work has stimulated a To supplement the existing data base, chromosome systematic survey of genome sizes (summarized in En counts were obtained for three species of Cienfue drizzi et al., 1985; Wendel et al., 1999). These studies gosia, two species of Thespesia,andL. kokioides. demonstrate a 3.5-fold range in genome sizes among Genome sizes were estimated for all of the above. For closely related diploid species in this single genus, rooting phylogenetic trees, Malva sylvestris L. was with values ranging from a low of 2.0 pg per 2C nu included as an outgroup. cleus for the ‘D-genome’ species from the New World to a high of approximately 7.0 pg per 2C nucleus Phylogenetic analysis for the ‘K-genome’ taxa from NW Australia. Gen ome sizes are relatively homogeneous within genome An approximately 1 kb segment of the 51 end of cel groups (summarized in Endrizzi et al., 1985; Wendel lulose synthase A1 (CesA1) was amplified and se et al., 1999) and each is also monophyletic (Wendel quenced as described in Cronn et al. (1999). This & Albert, 1992; Seelanan et al., 1997; Wendel et al., nuclear gene was selected because Southern hybrid 1999). The five allopolyploid species have genome ization surveys indicated that it was single-copy and sizes that are additive with respect to those of their because previous work established both experimental diploid progenitor genomes. protocols (conserved primer regions; amplification To further resolve the phylogenetic history of the conditions) and potential phylogenetic utility. The group and to better develop the organismal frame amplified portion spans exon 1 through exon 5 and 40 consists of approximately 60% exon nucleotide posi Results tions. Gaps caused few problems in aligning seq uences, which were readily aligned using ClustalW Phylogenetic analysis of CesA1 (Johnson et al., 1994). The aligned matrix was sub jected to both maximum likelihood and maxi Sequence data were generated for an approximately ∗ mum parsimony analysis using PAUP 4.0b8 1 kb portion of the cellulose synthase gene CesA1,in (Swofford, 2001). Nucleotide alignments are available cluding 615 exon nucleotides from the first five exons at http://www.botany.iastate.edu/∼jfw/HomePage/ and between 381 and 460 intron positions from introns jfwdata_sets.html. CesA1 sequences have been depos 1 through 4. Absolute lengths varied among taxa from ited in GenBank under accession numbers AF139442, 996 bp in the outgroup, M. sylvestris, to 1075 bp in G. AF139444, AF201886, AF376040, and AF376042– kirkii. The aligned data matrix contained 1146 nuc AF376048. Parsimony analysis was conducted us leotide positions, including 67 parsimony-informative ing a branch and bound search strategy with equal characters. weighting of all character-state transformations and Maximum parsimony analysis led to the recovery alignment gaps treated as missing data. Support for of a single tree (Figure 1) that is fully resolved and resolved clades was evaluated by jackknife resampling has high internal consistency with respect to the data (Farris, 1996) using 1000 replications and branch and (consistency index = 0.92; retention index = 0.81). bound searches. Maximum likelihood analysis was As shown in Figure 1, most clades are relatively performed using the HKY85 model of sequence evol strongly supported, with all but one being supported ution and empirically derived estimates of base fre by five or more character-state changes and having quencies (A = 0.29; C = 0.17; G = 0.21; T = 0.33), jackknife support greater than 75%. The topology transition/transversion rates (1.55), and the gamma shown extends and corroborates the initial phylogen shape parameter (1.56). etic scaffolding for the tribe Gossypieae based on the chloroplast gene ndhF and nuclear ribosomal ITS se Genome size estimation quences (Seelanan et al., 1997). As expected from the earlier analyses, Kokia and Gossypioides are sister- Newly expanded leaves from greenhouse-grown genera, these two being collectively sister to Gos plants were manually diced to release nuclei as de sypium. This clade of three genera is resolved as sister scribed (Johnston et al., 1999). In all cases, a leaf to a Thespesia + Hampea + Lebronnecia clade, with of the internal standard Pisum sativum cv. Minerva the latter clade additionally supported by a synapo Maple was included in nuclear isolation. Chopped morphic one-codon deletion (lysine) in exon 4 (the leaves were filtered through a 53 µm nylon mesh and only indel in the coding region of the CesA1 data propidium iodide was added to a final concentra set). The internal structure of the taxa included in the tion of 50 ppm. The mean fluorescence of nuclei was Thespesia + Hampea + Lebronnecia grouping is not quantified using a Coulter Epics Elite (Coulter Elec robustly resolved, as there is only weak support for the tronics, Hialeah FL) flow cytometer equipped with basal separation of H. appendiculata from the clade a water-cooled laser tuned at 514 nm and 500 mW. composed of L. kokioides, T. lampas,andT. thespe Fluorescence at >615 nm was detected with a pho sioides. Another point of agreement with the earlier tomultiplier screened by a long pass filter. Mean analysis of Seelanan et al. (1997) is that Thespesia 2C DNA content of each target species was calcu is paraphyletic or polyphyletic, suggesting once again lated by comparing its mean fluorescence with the the need for a more thorough study of this group. Fi mean fluorescence of the standard pea nuclei, which nally, the CesA1 data clearly and strongly support the has a 2C value of 9.56 pg DNA (Johnston et al., basal separation of Cienfuegosia from the remainder 1999). of the tribe, a position not unexpected given its un Chromosome counts usual chromosome numbers (Table 1) and divergent morphology (Fryxell, 1979). Chromosome counts were obtained from enzymatic Maximum likelihood analysis yielded the same to ally-digested root tips, pretreated in a saturated pology as that recovered from parsimony analysis, aqueous solution of alpha bromonaphthaline for indicating that the topology is unaffected by choice 90 min, as described by (Hanson et al., 1996) and of phylogeny reconstruction method. As shown by (Jewell & Islam-Faridi, 1994). the branch lengths in Figure 1, there was little het 41

Figure 1. Genome size evolution in the cotton tribe (Gossypieae). The shortest tree recovered from parsimony analysis, inferred from CesA1 sequence data, is identical in topology to the maximum likelihood phylogeny shown here. The number of character-state changes and jackknife support (%) from maximum parsimony analysis are shown above each internal branch, except for the T. lampas through L. kokioides clade, which was supported only by a single character (in addition to a one-codon deletion). The tree was rooted with the outgroup M. sylvestris (ingroup-outgroup branch length = 0.12). Genome sizes are shown at branch tips before species names, which are followed by somatic chro mosome numbers. Ancestral genome sizes were estimated using sum-of-squared-changes parsimony analysis (Maddison, 1991), a generalized least squared method (Martins & Hansen, 1997), and linear (= Wagner) parsimony (Swofford & Maddison, 1987). These three estimates are shown in boxes on the internal branches, with the ﬁrst two on the top line and the linear parsimony reconstructions on the bottom line (including the full range from minimum to maximum in cases where there was a range of equally parsimonious reconstructions). Inferred genome size increases are shown by shaded branches, decreases are indicated by un-ﬁlled branches, and ambiguities or stasis are denoted by hatched branches.

erogeneity in evolutionary rates for CesA1 among These data demonstrate a nearly sevenfold variation in the taxa studied. To formally evaluate this sugges- DNA content between the largest (T. populnea;2C= tion, we used the Kishino–Hasegawa likelihood ratio 8.2 pg) and smallest (G. kirkii and K. drynarioides, test to evaluate whether the CesA1 sequences ex- each with 2C = 1.2 pg) genomes measured. Whereas hibited clocklike behavior; log likelihoods with and relatively little variation was observed among the without clock enforcement were not signiﬁcantly dif- three Cienfuegosia species measured, the two Gos ferent (-ln L, no clock = 3108.03; molecular clock en- sypium species vary nearly twofold in DNA content forced = 3113.26; X2 = 10.459; p = 0.16), consistent and the three species of Thespesia exhibit greater than with equal evolutionary rates. a 2.5-fold difference in genome sizes.

Genome size estimation Chromosome counts

Flow cytometry was used to measure DNA content in To verify that the observed genome size variation re- all taxa included in the phylogenetic analysis. Replic- ﬂects changes at the same ploidy level, that is, that ate measurements from individual species were highly they do not result from polyploidy, chromosomes repeatable with low standard errors, and internal con- counts were obtained for Cienfuegosia tripartita trols yielded the expected genome sizes (data not (2n = 20), C. hitchcockii (2n = 20), C. yucatanen shown). Measured 2C values are shown in Table 1. sis (2n = 22), T. populnea (2n = 26), T. thespesioides 42

(2n = 26), and L. kokioides (2n = 26). These data, in possible reconstructions lead to the conclusion that conjunction with counts reported elsewhere (Fryxell, when summed across the phylogeny, the frequency 1979), conﬁrm that all plants included in the study are of 2C DNA content decrease actually exceeds that of diploid (Table 1). Accordingly, polyploidy cannot be increase. This result holds irrespective of any reason invoked to explain the dramatic genome size variation able adjustment of the phylogeny, which, as discussed among the taxa included in the study. above, is relatively robust.

Directionality of genome size change Discussion

The phylogenetic framework provided by Figure 1 Bidirectional evolution of genome size permits an heuristic evaluation of the directionality of genome size evolution in the tribe. To accomplish Recent years have witnessed an explosion in investig this we superimposed the DNA content estimates on ations of plant genome structure, culminating recently the phylogeny of Figure 1 and used three methods for in the near-complete sequencing of the Arabidopsis inferring ancestral states of continuously varying char genome (The Arabidopsis Genome Initiative, 2000). acters: (1) linear or Wagner parsimony (Swofford & This effort has led to a greatly enhanced understand Maddison, 1987), as implemented in PAUP∗;(2)sum ing of the important role that transposable elements, of-squared-changes parsimony analysis (Maddison, particularly retrotransposable elements, play in plant 1991), also as implemented in PAUP∗; and (3) a gener genome evolution (Wessler et al., 1995; Wessler, alized least squared method (Martins & Hansen, 1997) 1998; Bennetzen, 2000; Federoff, 2000). Especially implemented using the computer program COMPARE significant has been the realization that retroelements (Martins, 1999). These methods make no assumptions may proliferate on a massive scale in plant genomes about the relative likelihood of genome size contrac (Bennetzen, 1996; SanMiguel et al., 1996; Tikhonov tion or expansion, but are ‘agnostic’ (Bennetzen & et al., 1999; Bennett et al., 2000), in the process Kellogg, 1997a) in this respect. Because there may forming retroelement ‘landing pads’ that over evol be many equally parsimonious reconstructions under utionary time lead to a hierarchical archeology of linear parsimony, the full range of values was used to nested retroelements (SanMiguel et al., 1998). The explore the directionality of genome size change. notion that emerges from this literature is that plant As shown in Figure 1, all analytical methods lead genomes experience an inexorable increase in ge to the inference that genome size has fluctuated widely nome size over evolutionary time as a consequence during the radiation of the Gossypieae, and that both of retroelement accumulation. Superimposed on this genome size expansion and contraction have occurred process is a second phenomenon acting in the same repeatedly. Genome size is unambiguously inferred to direction, namely, polyploidy. Numerous comparat have decreased at least four times, once each in the ive genomic studies have increased our awareness of ancestors of G. raimondii, C. tripartita, the common the prevalence of polyploidy in plants, as well as its ancestor of T. lampas and T. thespesioides,andthe episodic and cyclical nature (Soltis & Soltis, 1999; common ancestor of G. kirkii and K. drynarioides. Otto & Whitton, 2000; Paterson et al., 2000; Wendel, Similarly, genome size is suggested to have increased 2000). in the lineages leading to G. herbaceum, H. appen Recognition of the dual processes of retroelement diculata,andT. populnea. In some cases it is unclear accumulation and polyploidization lends prima facie when these genome size changes took place. For ex credibility to the hypothesis that plant genomes ex ample, it is possible that the relatively large genome pand over evolutionary time, particularly in the ab of H. appendiculata arose is its immediate ancestor sence of counteracting mechanisms. Indeed, there (as shown in Figure 1), or earlier in the evolution of may be a coarse-grained trend toward genome ex the genus (only one species was sampled), or even pansion on a global phylogenetic scale in flowering before the divergence of Hampea from Thespesia and plants (Leitch et al., 1998) or perhaps within indi Lebronnecia. This last scenario, suggested as a possib vidual families, such as the grasses (Bennetzen & ility only by an extreme value from linear parsimony Kellogg, 1997a), or genera, such as the orchid genus analysis, would entail at least two genome size con Paphiopedilum (Cox et al., 1998). The phylogenetic tractions, once each in the ancestor of Lebronnecia evidence gathered to date, however, is circumstantial and Thespesia. Regardless of these uncertainties, all and subject to alternative interpretations. Nonetheless, 43 the hypothesis of unidirectional genome size evolu fashion that their sum of squared changes has been tion is inherently a phylogenetic statement, and hence minimized, or that the total evolutionary change has phylogenetic analysis represents the first and essen been minimized, regardless of the estimation method tial step in critically testing the hypothesis for any or its precision; mechanisms of genome size change particular plant group. do not lend themselves to such simplistic and regular Here we have provided a well-supported phylogen linearity. Nonetheless, and notwithstanding the ines etic framework for cotton and its allies, upon which we capable inaccuracy in the details of the quantitative superimposed genome sizes to document their phylo models, permitting the present distribution of gen genetic distribution. Polyploidy has been removed as ome sizes to reflect both expansion and contraction a possible confounding factor because we have shown is less assumption-laden and less restrictive than the that all of the included taxa are diploid. Hence, the alternative, increase-only scenario. present distribution of genome sizes reflects either If one accepts that the ancestral genome size re genome size increase or decrease, or perhaps both, construction methodology is conceptually legitimate since these genera and species diverged from their even if only quantitatively approximate, then one may common ancestor. As discussed by Bennetzen and use the phylogeny to assess the relative frequency of Kellogg (1997a), it is possible to invoke an ‘increase genome size increase and decrease. From Figure 1 only’ scenario, whereby genomes are prohibited from it is evident that it is unlikely that there has been significantly shrinking. If one adopts this strong as a unidirectional and parallel genome size expansion sumption, the ancestor of the entire tribe is inferred during the evolutionary history of the tribe. Instead, to have had a genome size equal to or smaller than several genome size expansions and contractions are the smallest one known among extant species (in the implicated. It seems likely, for example, that episodes present case, 2C = 1.2 pg). A corollary of this assump of genome contraction have occurred in the ancestry tion is that genome size expansion must have occurred of Kokia + Gossypioides, Cienfuegosia, Lebronnecia, in parallel in all lineages, with the possible exception Gossypium,and T. lampas + T. thespesioides. Equally of the one leading to Gossypioides and Kokia.Un probable are genome size expansions in T. popul der this scenario, the remarkable variation in extant nea and Hampea and Gossypium.InGossypium the genome sizes would reflect variation only in the rate largest genomes occur in a phylogenetically derived of genome size growth. Indeed, phylogenetic analysis (Seelanan et al., 1999) group of species from trop by itself may be incapable of providing a rigorous ical NW Australia; these species, with 2C contents refutation of this possibility, as the increase-only scen approximating 7.0 pg (J.McD. Stewart, pers. comm.), ario may be invoked irrespective of any particular would clearly be diagnosed as representing an addi distribution of extant and inferred ancestral genome tional example of genome size expansion. Similarly, sizes. For the reasons described in the following para sampling additional species in the other polytypic gen graphs, however, we view it as more plausible that era (Hampea, Thespesia,andCienfuegosia ) would both expansion and contraction have occurred. likely reveal additional genome size variation and If one conceptually permits both genome ex hence episodes of expansion and contraction. pansion and contraction, thereby imposing a less Our results for the Gossypieae provide a phylogen stringent assumption on the origin of the present etic suggestion of episodic genome size contraction, distribution of genome sizes, one may employ widely and thereby contribute to an increasing recognition applied methods (Swofford & Maddison, 1987; Mad that ‘obese’ plant genomes may occasionally go ‘on dison, 1991; Bennetzen & Kellogg, 1997a; Martins a diet’ (Rabinowicz, 2000). Much of the evidence & Hansen, 1997; Cunningham et al., 1998; Martins, for this dieting is circumstantial but collectively the 1999; Oakley & Cunningham, 2000) to infer ances data are compelling. First, our results for the Gos tral genome sizes. Of course, one might challenge the sypieae are writ large on a phylogenetic scale: plants quantitative details of the ancestral state reconstruc possessing small or large genomes are phylogenetic- tion methodologies, which do not necessarily model ally widespread throughout the angiosperms (Leitch biological processes in a realistic manner and which et al., 1998; Stace, 2000). Genome size data exist for are subject to various biases and large confidence only about 1.4% of angiosperm species (Bennett et al., intervals (Cunningham et al., 1998; Oakley & Cun 1997; Bennett et al., 2000), yet the available data ningham, 2000; Polly, 2001). For example, it seems base documents remarkable variation in genome size unlikely that genome sizes have evolved in such a within numerous genera and families. A. thaliana is 44 often cited as having the smallest angiosperm genome, et al., 1998; Sossey-Alaouni et al., 1998; Brubaker although Cardamine amara from the same family et al., 1999; Wilson et al., 1999; Paterson et al., 2000). (Brassicaceae) has a genome size half again as small Even Arabidopsis appears to have experienced one and (2C = 0.11; Bennett & Smith, 1991). The distantly possible several cycles of polyploidy during its evol related horse-chestnut, Aesculus hippocastanum (Hip utionary history (The Arabidopsis Genome Initiative, pocastanaceae sensu strictu, or Sapindaceae sensu 2000; Grant et al., 2000; Vision et al., 2000), yet its lato), has a genome as small as that of Arabidop genome usually is considered a model of streamlin sis, even though it clearly is paleopolyploid (2n = 40). ing. Clearly then, the streamlining process has entailed In a recent phylogenetic analysis of angiosperm gen massive loss of redundant genomic sequence. ome sizes (Leitch et al., 1998) small genomes were A third body of evidence pertinent to the possibil evident in all orders and larger genomes were widely ity of genome size contraction stems from recent stud distributed as well. The range of variation in C-values ies that increased our awareness of the significance is remarkable within phylogenetically disparate lin of deletional mechanisms during genome evolution eages, including the Asterids (0.3–24.8pg), Rosids (Petrov, 2001). These mechanisms are many and (0.1–16.5pg), Ranunculales (0.5–25.1pg), and mono- varied, and sometimes the specific molecular events cots (0.3–127.4pg). These data may only be recon responsible are rather poorly understood. In Dro ciled with unidirectional genome size evolution by the sophila, for example, examination of ‘dead-on- ar most stringent assumptions, including long-term re rival’ non-LTR retroelements revealed a spontaneous tention of tiny genomes in phylogenetically disparate rate of DNA loss many times higher than that ex lineages as well as rampant genome size expansion hibited by mammalian pseudogenes (Petrov et al., in others. A more parsimonious interpretation, in our 1996). Moreover, the efficiency of this mechanism view, is that the foregoing data indicate that genome may correlate with genome size; Laupala crickets, size is evolutionary labile and subject to both increase with a genome size 11-fold higher than Drosophila, and decrease. It also seems likely, in our opinion, lose DNA through this process 40 times more slowly that patterns such as those reported here will be in (Petrov et al., 2000). Many of the deleted regions are creasingly observed as more genera and families are associated with short direct repeats, suggesting loss subjected to joint phylogenetic analysis and genome through recombination or replication slippage (Petrov size determinations. & Hartl, 1997). These studies were recently exten A second source of evidence bearing on the like ded to Podisma grasshoppers, which have a genome lihood of genome size contraction derives from com that is 10 times larger than Laupala; rates of deletion parative genomics, particularly comparative mapping in ‘dead- on-arrival’ nuclear pseudogenes are much studies (Paterson et al., 2000), which have informed us lower in the former than the latter, relative to rates of about the ‘deep history’ of angiosperm genomes. As point substitution (Bensasson et al., 2001). Rates of noted earlier, polyploidy is an active, ongoing process insertion or deletion in introns also may correlate with in angiosperms (Masterson, 1994; Leitch & Bennett, genome size, such that animals with larger genomes 1997; Soltis & Soltis, 1999; Otto & Whitton, 2000; may possess larger introns (Hughes & Hughes, 1995; Wendel, 2000). Because genome doubling has been Ogata et al., 1996; Moriyama et al., 1998; Deutsch & continuing since angiosperms first appeared definit Long, 1999; Vinogradov, 1999). ively in the Cretaceous, most if not all angiosperm In plants, recent research involving experimentally genomes have experienced one or more episodes of induced double-stranded breaks showed that deletions polyploidization at various times in the past. Thus, were larger and insertions rarer in A. thaliana than in the modern angiosperm genome typically consists of a Nicotiana tabacum, which has a much larger genome series of nested duplications of varying antiquity, only (Kirik et al., 2000). Analysis of the recently published some of which descend to the present relatively un Arabidopsis genome sequence suggests that unequal scathed by subsequent evolutionary disruptions. This crossing over may lead to deletion of genes or gen paleopolyploidy often is revealed by detailed compar omic segments (The Arabidopsis Genome Initiative, ative mapping studies or other approaches (Reinisch 2000). Recombinational mechanisms are also likely et al., 1994; Moore et al., 1995; Lagercrantz & Ly important in genome size contraction. This has el diate, 1996; Shoemaker et al., 1996; Bennetzen & egantly been shown by the work of Schulman and Freeling, 1997; Gaut & Doebley, 1997; Gómez et al., collaborators using natural and experimental popula 1998; Kellogg, 1998; Lagercrantz, 1998; Muravenko tions of barley (Vicient et al., 1999; Kalendar et al., 45

2000; Shirasu et al., 2000). In one particular 66 kb changes, each with their own fitness consequences. region a minimum of 15 distinct retrotransposon in Thus, an important avenue for future research is to sertion events were identified, yet most long terminal simultaneously assess the directionality of change for repeats (LTRs) exist in single copy (Shirasu et al., multiple and different genomic components within a 2000). Presumably the responsible mechanism is ho well-understood phylogenetic framework, using mod mologous recombination between retroelement LTRs els like the one exemplified in the present study. It leading to internal deletions. An excess of solo LTRs may be, for example, that overall genome size change has been observed throughout the genus Hordeum primarily reflects copy-number increase or decrease (Vicient et al., 1999), and remarkably, correlations for one or several families of repetitive elements, between full-length elements and LTR copy number whose net effect quantitatively overwhelms other exist on local ecological scales in H. spontaneum from genomic components evolving in a countervailing dir Israel (Kalendar et al., 2000). The latter example un ection, each with their own effects on fitness. Altern derscores the possibility of rapid change in genome atively, one might envision cases where genome size size, mediated through homologous recombination. itself is particularly responsive to selection, thereby Perhaps larger deletions involving redundant chromo imposing a unidirectional response for most genomic some segments or blocks of repetitive DNA could constituents. At present this is relatively unexplored be deleted by similar mechanisms, as speculated by terrain in the field of genome evolution. By combining Bennetzen and Kellogg (1997a). the tools of phylogenetics and genomics, insights into these and related questions will undoubtedly soon be Concluding remarks forthcoming.

Much remains to be learned regarding deletional pro Acknowledgements cesses and their significance in genome evolution, but the foregoing attests to an increasing recognition We thank D. Petrov for helpful discussion and T. that plant genome sizes are evolutionary labile and Haselkorn and George Hodnett for technical assist that deletional mechanisms may play a more prom ance. This work was supported by the National Sci inent role than previously believed. This suggests, ence Foundation (USA) and the Texas Agric. Expt. in conjunction with the phylogenetic and compara Station. tive mapping evidence discussed above, that gen ome size contraction is a real and perhaps common evolutionary phenomenon. Our results for the single References Malvaceous tribe Gossypieae implicate one or more of these or other mechanisms of genome contraction, The Arabidopsis Genome Initiative, 2000. Analysis of the genome and hence contribute to the growing awareness that sequence of the flowering plant, Arabidopsis thaliana.Nature 408: 796–815. obese plant genomes frequently go ‘on a diet’ (Ra Bennett, M.D., 1985. Intraspecific variation in DNA amount and the binowicz, 2000). Accordingly, present genome sizes nucleotypic dimension in plant genetics, pp. 283–302 in Plant may most appropriately be viewed as reflecting a dy Genetics, edited by M. Freeling. A. R. Liss, NY. Bennett, M.D., 1987. Variation in genomic form and its ecological namic balance between opposing forces of expansion implications. New Phytol. 106: 177–200. and contraction, as recently outlined by Petrov (2001). Bennett, M.D., P. Bhandol & I.J. Leitch, 2000. Nuclear DNA The foregoing discussion highlights both the evol amounts in angiosperms and their modern uses – 807 new utionary lability of genome size and that the process estimates. Ann. Bot. 86: 859–909. Bennett, M.D., A.V. Cox & I.J. Leitch, 1997. Angiosperm DNA c- entails both expansion and contraction. This recogni values database. http://www.rbgkew.org.uk/cval/database1.html tion may constitute a step toward understanding the Bennett, M.D. & I.J. Leitch, 1995. Nuclear DNA amounts in evolutionary significance of genome size change, but angiosperms. Ann. Bot. 76: 113–176. perhaps this is only a rather small step. The pattern Bennett, M.D. & J.B. Smith, 1991. Nuclear DNA amounts in angiosperms. Phil. Trans. R. Soc. London B 334: 309–345. described is global in scope, that is, total DNA content Bennetzen, J.L., 1996. The contributions of retroelements to plant per cell, and while genome size itself may have bio genome organization, function and evolution. Trends Microbiol. logical significance or be visible to natural selection 4: 347–353. (Bennett, 1985; Bennett, 1987; Price, 1988; Gregory Bennetzen, J.L., 2000. Transposable element contributions to plant gene and genome evolution. Plant Mol. Biol. 42: 251–269. & Hebert, 1999; Petrov, 2001), it may be that DNA Bennetzen, J.L. & M. Freeling, 1997. The unified grass genome: content represents the net effect of myriad finer-scale synergy in synteny. Genome Res. 7: 301–306. 46

Bennetzen, J.L. & E.A. Kellogg, 1997a. Do plants have a one-way taneum)byBARE -1 retrotransposon dynamics in response to ticket to genomic obesity? Plant Cell 9: 1509–1514. sharp microclimatic divergence. Proc. Natl. Acad. Sci. USA 97: Bennetzen, J.L. & E.A. Kellogg, 1997b. Reply. Plant Cell 9: 1901– 6603–6607. 1902. Kellogg, E.A., 1998. Relationships of cereal crops and other grasses. Bensasson, D., D.A. Petrov, D.-X. Zhang, D.L. Hartl & G.M. He Proc. Natl. Acad. Sci. USA 95: 2005–2010. witt, 2001. Genomic gigantism: DNA loss is slow in mountain Kirik, A., S. Salomon & H. Puchta, 2000. Species-specific double- grasshoppers. Mol. Biol. Evol. 18: 246–253. strand break repair and genome evolution in plants. EMBO J. Brubaker, C.L., A.H. Paterson & J.F. Wendel, 1999. Compar 2000: 5562–5566. ative genetic mapping of allotetraploid cotton and its diploid LaDuke, J.C. & J.F. Doebley, 1995. A chloroplast DNA based progenitors. Genome 42: 184–203. phylogeny of the Malvaceae. Syst. Bot. 20: 259–271. Cox, A.V., G.J. Abdelnour, M.D. Bennett & I.J. Leitch, 1998. Lagercrantz, U., 1998. Comparative mapping between Arabidop Genome size and karyotype evolution in the slipper orchids sis thaliana and Brassica nigra indicates that Brassica genomes (Cypripedioideae: Orchidaceae). Amer. J. Bot. 85: 681–687. have evolved through extensive genome replication accompanied Cronn, R., R.L. Small & J.F. Wendel, 1999. Duplicated genes evolve by chromosome fusions and frequent recombinations. Genetics independently following polyploid formation in cotton. Proc. 150: 1217–1228. Natl. Acad. Sci. USA 96: 14406–14411. Lagercrantz, U. & D.J. Lydiate, 1996. Comparative genome map Cunningham, C.W., K.E. Omland & T.H. Oakley, 1998. Recon ping in Brassica. Genetics 144: 1903–1910. structing ancestral character states: a critical reappraisal. Trends Leitch, I.J. & M.D. Bennett, 1997. Polyploidy in angiosperms. Ecol. Evol. 13: 361–366. Trends Plant Sci. 2: 470–476. Deutsch, M. & M. Long, 1999. Intron-exon structure of eukaryotic Leitch, I.J., M.W. Chase & M.D. Bennett, 1998. Phylogenetic anal model organisms. Nucl. Acids Res. 27: 3219–3228. ysis of DNA C-values provides evidence for a small ancestral Endrizzi, J.E., E.L. Turcotte & R.J. Kohel, 1985. Genetics, cytogen genome size in flowering plants. Ann. Bot. (suppl. A) 82: 85–94. etics, and evolution of Gossypium. Adv. Genet. 23: 271–375. Maddison, W.P., 1991. Squared-change parsimony reconstructions Farris, J.S., 1996. Parsimony jackknifing outperforms neighbor- of ancestral states for continuous-valued characters on a phylo joining. Cladistics 12: 99–124. genetic tree. Syst. Zool. 40: 304–314. Federoff, N., 2000. Transposons and genome evolution in plants. Martins, E.P., 1999. COMPARE, Version 4.3. Computer Programs Proc. Natl. Acad. Sci. USA 97: 7002–7007. for the Statistical Analysis of Comparative Data. Distributed Fryxell, P.A., 1979. The Natural History of the Cotton Tribe. Texas by the author via the WWW at http://darkwing.uoregon.edu/ A&M University Press, College Station, TX. ∼compare4/. Department of Biology, University of Oregon, Fryxell, P.A., 1992. A revised taxonomic interpretation of Gos Eugene, Oregon. sypium L. (Malvaceae). Rheedea 2: 108–165. Martins, E.P. & T.F. Hansen, 1997. Phylogenies and the compar Gaut, B.S. & J.F. Doebley, 1997. DNA sequence evidence for the ative method: a general approach to incorporating phylogenetic segmental allotetraploid origin of maize. Proc. Natl. Acad. Sci. information into the analysis of interspecific data. Amer. Nat. USA 94: 6808–6814. 149: 646–667. Gómez, M.I., M.N. Islam-Faridi, M.S. Zwick, D.G. Czeschin, G.E. Masterson, J., 1994. Stomatal size in fossil plants: evidence for Hart, R.A. Wing, D.M. Stelly & H.J. Price, 1998. Tetraploid polyploidy in majority of angiosperms. Science 264: 421–424. nature of Sorghum bicolor (L.) Moench. J. Hered. 89: 188–190. Miege, J. & N. Josserand, 1972. Nombres chromosomiques Grant, D., P. Cregan & R.C. Shoemaker, 2000. Genome organiz d’especes africaines et malgaches. Candollea 27: 283–292. ation in dicots: genome duplication in Arabidopsis and synteny Moore, G., K.M. Devos & Z. Wang, 1995. Grasses, line up and form between soybean and Arabidopsis. Proc. Natl. Acad. Sci. USA a circle. Curr. Biol. 5: 737–739. 97: 4168–4173. Moriyama, E.N., D.A. Petrov & D.L. Hartl, 1998. Genome size and Gregory, T.R. & P.D.N. Hebert, 1999. The modulation of DNA con intron size in Drosophila. Mol. Biol. Evol. 15: 770–773. tent: proximate causes and ultimate consequences. Genome Res. Muravenko, O., A.R. Fedotov, E.O. Punina, L.I. Federova, V.G. 9: 317–324. Grif & A.V. Zelenin, 1998. Comparison of chromosome BrdU Hanson, R.E., M.N. Islam-Faridi, E.A. Percival, C.F. Crane, Y. Ji, Hoechst-Giemsa banding patterns of the A1 and (AD)2 genomes T.D. McKnight, D.M. Stelly & H.J. Price, 1996. Distribution of of cotton. Genome 41: 616–625. 5S and 18S-28S rDNA loci in a tetraploid cotton (Gossypium hir Oakley, T.H. & C.W. Cunningham, 2000. Independent contrasts sutum L.) and its putative diploid ancestors. Chromosoma 105: succeed where ancestor reconstruction fails in a known bacterio 55–61. phage phylogeny. Evolution 54: 397–405. Hughes, A.L. & M.K. Hughes, 1995. Small genomes for better Ogata, H., W. Fujibuchi & M. Kanehisa, 1996. The size differences flyers. Nature 377: 391. among mammalian introns are due to the accumulation of small Jewell, D.C. & M.N. Islam-Faridi, 1994. Details of a technique for deletions. FEBS Lett. 390: 99–103. somatic chromosome preparation and C-banding of maize, pp. Otto, S.P. & J. Whitton, 2000. Polyploid incidence and evolution. 484–493 in The Maize Handbook, edited by M. Freeling & V. Annu. Rev. Genet. 34: 401–437. Walbot. Springer, New York. Paterson, A.H., J.E. Bowers, M.D. Burow, X. Draye, C.G. Elsik, Johnson, J.D., D.G. Higgins & T.J. Gibson, 1994. CLUSTAL W: C.-X. Jiang, C.S. Katsar, Y.-R. Lan, R. Ming & R.J. Wright, improving the sensitivity of progressive multiple sequence align 2000. Comparative genomics of plant chromosomes. Plant Cell ment through sequence weighting, position specific gap penalties 12: 1523–1539. and weight matrix choice. Nucl. Acids Res. 22: 4673–4680. Petrov, D.A., 1997. Slow but steady: reduction to genome size Johnston, J.S., M.D. Bennett, A.L. Rayburn, D.W. Galbraith & through biased mutation. Plant Cell 9: 1900–1901. H.J. Price, 1999. Reference standards for determination of DNA Petrov, D.A., 2001. Evolution of genome size: new approaches to content of plant nuclei. Amer. J. Bot. 86: 609–613. an old problem. Trends Genet. 17: 23–28. Kalendar, R., J. Tanskanen, S. Immonen, E. Nevo & A.H. Schul- Petrov, D.A. & D.L. Hartl, 1997. Trash DNA is what gets thrown man, 2000. Genome evolution in wild barley (Hordeum spon away: high rate of DNA loss in Drosophila. Gene 205: 279–289. 47

Petrov, D.A., E.R. Lozovskaya & D.L. Hartl, 1996. High intrinsic Stace, C.A., 2000. Cytology and cytogenetics as a fundamental rate of DNA loss in Drosophila. Nature 384: 346–349. taxonomic resource for the 20th and 21st century. Taxon 49: Petrov, D.A., T.A. Sangster, J.S. Johnston, D.L. Hartl & K.L. Shaw, 451–477. ∗ 2000. Evidence for DNA loss as a determinant of genome size. Swofford, D.L., 2001. PAUP : Phylogenetic Analysis Using Parsi ∗ Science 287: 1060–1062. mony ( and Other Methods), Version 4. Sinauer Associates, Polly, P.D., 2001. Paleontology and the comparative method: an Sunderland, MA. cestral node reconstructions versus observed node values. Amer. Swofford, D.L. & W.P. Maddison, 1987. Reconstructing ancestral Nat. 157: 596–609. character states under Wagner parsimony. Math. Biosci. 87: 199– Price, H.J., 1988. Nuclear DNA content variation within angiosperm 229. species. Evol. Trends Plants 2: 53–60. Thomas, C.A., 1971. The genetic organisation of chromosomes. Rabinowicz, P.D., 2000. Are obese plant genomes on a diet? Annu. Rev. Genet. 5: 237–256. Genome Res. 10: 893–894. Tikhonov, A., P. SanMiguel, Y. Nakajima, N.M. Gorenstein, J.L. Reinisch, A.J., J. Dong, C.L. Brubaker, D.M. Stelly, J.F. Wendel Bennetzen & Z. Avramova, 1999. Colinearity and its exceptions & A.H. Paterson, 1994. A detailed RFLP map of cotton, Gos in orthologous adh regions of maize and sorghum. Proc. Natl. sypium hirsutum × G. barbadense: chromosome organization Acad. Sci. USA 96: 7409–7414. and evolution in a disomic polyploid genome. Genetics 138: Vicient, C.M., A. Suoniemi, K. Anamthawat-Jonsson, J. Tanskanen, 829–847. A. Beharav, N.E. & A.H. Schulman, 1999. Retrotransposon SanMiguel, P., B.S. Gaut, A. Tikhonov, Y. Nakajima & J.L. Ben BARE-1 and its role in genome evolution in the genus Hordeum. netzen, 1998. The paleontology of intergene retrotransposons of Plant Cell 11: 1769–1784. maize: dating the strata. Nat. Gen. 20: 43–45. Vinogradov, A.E., 1999. Intron-genome size relationship on a large SanMiguel, P., A. Tikhonov, Y.K. Jin, N. Motchoulskaia, D. Za evolutionary scale. J. Mol. Evol. 49: 376–384. kharov, A. Melake-Berhan, P.S. Springer, K.J. Edwards, M. Lee, Vision, T.J., D.G. Brown & S.D. Tanksley, 2000. The origins of Z. Avramova & J.L. Bennetzen, 1996. Nested retrotransposons genomic duplications in Arabidopsis. Science 290: 2114–2117. in the intergenic regions of the maize genome. Science 274: Wendel, J.F., 2000. Genome evolution in polyploids. Plant Mol. 765–768. Biol. 42: 225–249. Seelanan, T., C.L. Brubaker, J.M. Stewart, L.A. Craven & J.F. Wendel, J.F. & V.A. Albert, 1992. Phylogenetics of the cotton Wendel, 1999. Molecular systematics of Australian Gossypium genus (Gossypium L.): character-state weighted parsimony anal section Grandicalyx (Malvaceae). Syst. Bot. 24: 183–208. ysis of chloroplast DNA restriction site data and its systematic Seelanan, T., A. Schnabel & J.F. Wendel, 1997. Congruence and and biogeographic implications. Syst. Bot. 17: 115–143. consensus in the cotton tribe. Syst. Bot. 22: 259–290. Wendel, J.F., R.L. Small, R.C. Cronn & C.L. Brubaker, 1999. Shirasu, K., A.H. Schulman, T. Lahaye & P. Schulze-Lefert, 2000. Genes, jeans, and genomes: reconstructing the history of cotton. A contiguous 66-kb barley DNA sequence provides evidence for Plant evolution in man-made habitats, pp. 133–161 in Proc. VIIth reversible genome expansion. Genome Res. 10: 908–915. Int. Symp. Int. Org. Plant Biosystemat. Amsterdam. Shoemaker, R.C., K. Polzin, J. Labate, J. Specht, E.C. Brummer, Wessler, S.R., 1998. Transposable elements associated with normal T. Olson, N. Young, V. Concibido, J. Wilcox, J.P. Tamulonis, G. plant genes. Physiol. Plantarum 103: 581–586. Kochert & H.R. Boerma, 1996. Genome duplication in soybean Wessler, S.R., T.E. Bureau & S.E. White, 1995. LTR-retro (Glycine subgenus soja). Genetics 144: 329–338. transposons and MITEs: important players in the evolution of Soltis, D.E. & P.S. Soltis, 1999. Polyploidy: origins of species and plant genomes. Curr. Opin. Genet. Dev. 5: 814–821. genome evolution. Trends Ecol. Evol. 9: 348–352. Wilson, W.A., S.E. Harrington, W.L. Woodman, M. Lee, M.E. Sor Sossey-Alaouni, K., H. Serieys, M. Tersac, P. Lambert, E. Schilling, rells & S.R. McCouch, 1999. Inferences on the genome structure Y. Griveau, F. Kaan & A. Bervill b,´ 1998. Evidence for several of progenitor maize through comparative analysis of rice, maize genomes in Helianthus. Theor. Appl. Genet. 97: 422–430. and the domesticated Panicoids. Genetics 153: 453–473.